30 Claude Prompts for a Methods Section You Can Defend

<context> I will paste access notes and, if I have it, a planned or known headcount. Describe only that sample. Do not complete a missing n from a typical paper in my field. </context> <inputs> - Research question: [RESEARCH QUESTION] - Planned or known n: [SAMPLE N OR UNKNOWN] - Sample or access notes: [PASTE SAMPLE NOTES] - Inclusion rule: [INCLUSION RULE] - Setting: [SETTING] - Who I could not reach: [MISSING POPULATION] </inputs> <task> Write a participants subsection: who is in the set, how they entered, who is missing, and the n I supplied. If [SAMPLE N OR UNKNOWN] is UNKNOWN, keep every sentence n-free and list the cells I must open in the file. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent n, a response rate, or a demographic split I did not paste. - If a cell is missing, write NOT IN PASTE and mark UNVERIFIED. - Do not upgrade a convenience sample into a representative one. </constraints> <format> Participants prose, then a table of known versus NOT IN PASTE fields, then UNVERIFIED rows. </format>

<context> I need inclusion and exclusion criteria as I actually applied them. Do not add a criterion that would make the sample look cleaner. </context> <inputs> - Research question: [RESEARCH QUESTION] - Inclusion rule: [INCLUSION RULE] - Exclusion rule: [EXCLUSION RULE OR NONE] - Protocol notes: [PROTOCOL NOTES] - Borderline cases I already decided: [BORDERLINE CASES OR NONE] </inputs> <task> Write inclusion and exclusion as numbered rules a second reader could apply. For each rule: the test, where it sits in the protocol, and one borderline I already decided if I pasted one. Flag any rule that appears only in my memory and not in the paste as UNVERIFIED. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent an exclusion that would tidy a messy sample. - Do not invent n excluded or a flow-diagram count. - If [EXCLUSION RULE OR NONE] is NONE, do not invent exclusions. - If a rule needs a source or form I did not paste, mark it UNVERIFIED. </constraints> <format> Numbered inclusion rules, numbered exclusion rules, borderline notes, UNVERIFIED rows. </format>

<context> I need a recruitment subsection that matches how people or records entered. Do not invent a channel, an incentive, or a yield I did not paste. </context> <inputs> - Research question: [RESEARCH QUESTION] - Recruitment notes: [PASTE RECRUITMENT NOTES] - Channels I actually used: [CHANNELS] - Incentive if any: [INCENTIVE OR NONE] - Yield if I counted it: [YIELD OR UNKNOWN] </inputs> <task> Write recruitment as run: channels, who invited, what they were told, incentive only if I named one, and yield only if I pasted it. If [YIELD OR UNKNOWN] is UNKNOWN, keep every sentence count-free and list the logs I must open. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a flyer, a clinic, a panel, or a payment I did not name. - Do not invent a response rate or a refused-n. - If a step is missing, write NOT IN PASTE and mark UNVERIFIED. </constraints> <format> Recruitment prose, then a channel checklist with NOT IN PASTE tags, then UNVERIFIED rows. </format>

<context> This study uses a corpus, archive, or record set rather than (or as well as) living participants. Describe only access I pasted. Do not invent a collection I cannot open. </context> <inputs> - Research question: [RESEARCH QUESTION] - Access notes: [PASTE ACCESS NOTES] - What is in the set: [CORPUS DESCRIPTION] - Dates of records if known: [RECORD DATES OR UNKNOWN] - Who I cannot reach: [MISSING POPULATION] </inputs> <task> Write a corpus or records subsection: what the set is, how I gained access, what is excluded, and date bounds I supplied. If [RECORD DATES OR UNKNOWN] is UNKNOWN, keep sentences date-free and list finding-aid cells I must open. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent an archive, a database, a license, or n records. - Do not upgrade a convenience scrape into a complete census. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Corpus prose, then known versus NOT IN PASTE fields, then UNVERIFIED rows. </format>

<context> My sample table is incomplete. I need methods sentences that stay honest about missing cells. Do not fill n, percents, or demographics from a typical paper. </context> <inputs> - Research question: [RESEARCH QUESTION] - Sample notes or table: [PASTE SAMPLE NOTES] - Planned or known n: [SAMPLE N OR UNKNOWN] - Cells I know are empty: [EMPTY CELLS] - Who I could not reach: [MISSING POPULATION] </inputs> <task> Write a missing-cells memo: which fields are known, which are NOT IN PASTE, which sentences must stay n-free, and a participants paragraph that uses only known cells. Leave every count as UNKNOWN unless I pasted it. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent n, a percent, an age mean, or a gender split. - Do not round a missing cell into a tidy table. - If [SAMPLE N OR UNKNOWN] is UNKNOWN, refuse a completed flow sentence and mark UNVERIFIED. </constraints> <format> Known versus UNKNOWN table, then an n-free paragraph, then UNVERIFIED rows. </format>

<context> I need an instruments subsection from tools I actually named. Do not invent a scale, an item, a Cronbach alpha, or a version I did not paste. </context> <inputs> - Research question: [RESEARCH QUESTION] - Instruments I used: [INSTRUMENTS] - Protocol notes: [PROTOCOL NOTES] - Versions or editions if I have them: [VERSIONS OR UNKNOWN] - Citation style: [CITATION STYLE] - Instrument source I already have (optional): [PASTE INSTRUMENT SOURCE OR NONE] </inputs> <task> Describe each item in [INSTRUMENTS]: what it captured, how I administered it, and version only if I pasted one. Cite only labels that appear in [PASTE INSTRUMENT SOURCE OR NONE]. If a version, alpha, or item count is missing, write UNKNOWN and mark UNVERIFIED. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a validated scale, a cutoff, or a reliability coefficient. - Do not invent author-year for an instrument I did not paste. - If I wrote my own items, say so. Do not dress them as a published measure. </constraints> <format> One block per instrument, then UNKNOWN fields, then UNVERIFIED rows. </format>

<context> I will paste what I actually did in the field or at the desk. The procedure must match that paste, including awkward steps. Do not add a visit, a recorder, or a pilot I did not name. </context> <inputs> - Research question: [RESEARCH QUESTION] - What I actually did: [WHAT I ACTUALLY DID] - Planned protocol if it differs: [PROTOCOL OR SAME] - Word limit: [WORD COUNT] - Citation style: [CITATION STYLE] </inputs> <task> Rewrite [WHAT I ACTUALLY DID] as a procedures subsection a second reader could follow: order of tasks, who was present, what was collected, and how long a session lasted only if I pasted a duration. Stay under [WORD COUNT]. List deviations from [PROTOCOL OR SAME] separately. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a consent script, a recorder, or a second interviewer. - If duration, n sessions, or site count is missing, write UNKNOWN and mark UNVERIFIED. - If the paste is a plan and not a log, label the draft AS PLANNED. </constraints> <format> Numbered procedures under [WORD COUNT], then AS PLANNED versus AS RUN, then UNVERIFIED rows. </format>

<context> I need timing, sites, and waves reported as I ran them. Do not invent a second site or a follow-up wave that would make the design look stronger. </context> <inputs> - Research question: [RESEARCH QUESTION] - Protocol notes: [PROTOCOL NOTES] - Sites I actually used: [SITES] - Dates or waves: [DATES OR WAVES OR UNKNOWN] - What I skipped: [SKIPPED WAVES OR NONE] </inputs> <task> Write a timing and sites subsection: which sites, which waves, which dates I supplied, and what I skipped. If [DATES OR WAVES OR UNKNOWN] is UNKNOWN, keep sentences count-free and list the calendar cells I must open. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a site, a wave, or n visits. - Do not invent that a skipped wave was optional if the protocol required it. Name the skip. - If a date is missing, write UNKNOWN and mark UNVERIFIED. </constraints> <format> Timing and sites prose, then a wave table with UNKNOWN cells, then UNVERIFIED rows. </format>

<context> Collection did not match the protocol in places I will name. Log those deviations. Do not invent extra ones to look transparent, and do not tidy away ones I pasted. </context> <inputs> - Research question: [RESEARCH QUESTION] - Planned protocol: [PROTOCOL] - What I actually did: [WHAT I ACTUALLY DID] - Deviations I already logged: [DEVIATIONS] - Ethics note if any: [ETHICS NOTE OR NONE] </inputs> <task> Write a collection-deviations subsection from [DEVIATIONS] plus any clash you can see between [PROTOCOL] and [WHAT I ACTUALLY DID]. For each: what changed, when if I said so, whether ethics already knew, and a methods sentence. Do not add a deviation that is not in the paste. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a missed visit, a broken recorder, or a protocol amendment. - If [ETHICS NOTE OR NONE] is NONE, do not invent that a board approved the change. - If a change needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Numbered deviations with methods sentences, then a not-in-paste clash list, then UNVERIFIED rows. </format>

<context> I need data storage and handling sentences that match what I actually used. Do not invent a platform, an encryption claim, or a retention year I did not paste. </context> <inputs> - Research question: [RESEARCH QUESTION] - Protocol notes: [PROTOCOL NOTES] - Where files live: [STORAGE] - Who can open them: [ACCESS RULE] - Retention if I know it: [RETENTION OR UNKNOWN] - Identifiers I kept or stripped: [IDENTIFIERS] </inputs> <task> Write a storage and handling subsection: where data sat, who could open it, whether identifiers stayed, and retention only if I pasted it. If [RETENTION OR UNKNOWN] is UNKNOWN, keep that sentence out and list the policy I must open. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a vendor, a lock, or n years of retention. - Do not invent that data were anonymised if I only said de-identified, or the reverse. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Storage prose, then a known versus UNKNOWN checklist, then UNVERIFIED rows. </format>

<context> I need an analysis subsection from steps I already named, not results. Do not invent a test, a model, or an n I did not name. </context> <inputs> - Research question: [RESEARCH QUESTION] - Design: [STUDY DESIGN] - Analysis notes: [PASTE ANALYSIS NOTES] - Software I can use: [SOFTWARE] - Sample n if known: [SAMPLE N OR UNKNOWN] - Word limit: [WORD COUNT] </inputs> <task> Write an analysis plan: units, preparation, primary analysis, secondary or exploratory work, and what I will not claim. Leave every count as UNKNOWN unless I pasted it. Stay under [WORD COUNT]. If the work already ran, label steps AS RUN only when the paste is a log. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent n, p-values, effect sizes, or a finding. - Do not add a test I cannot run in [SOFTWARE] unless you label it OPTIONAL, NEEDS NEW TOOL. - If [SAMPLE N OR UNKNOWN] is UNKNOWN, refuse a power claim and mark UNVERIFIED. </constraints> <format> Numbered analysis steps, then a will-not-claim list, then UNKNOWN cells, then UNVERIFIED rows. </format>

<context> I need the variables, codes, or materials I actually used listed so a reader can follow the analysis. Do not invent a measure, a code, or a codebook I did not paste. </context> <inputs> - Research question: [RESEARCH QUESTION] - Analysis notes or codebook excerpts: [PASTE ANALYSIS NOTES] - Outcomes I named: [OUTCOMES] - Predictors or themes I named: [PREDICTORS OR THEMES] - Citation style: [CITATION STYLE] </inputs> <task> List variables or codes from the paste only: name, role (outcome, predictor, covariate, theme, other), how I measured or applied it, and any missing definition. Quote codebook lines only if they appear in the paste. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a variable, a code, or a cutoff. - Do not invent n levels or n themes. - If a name is in [OUTCOMES] but not defined in the notes, mark UNVERIFIED, NEED DEFINITION. </constraints> <format> Table: name | role | definition in paste | UNVERIFIED. Then a NEED DEFINITION list. </format>

<context> I need a missingness plan that does not invent how much is missing. Use only rules I pasted. Do not invent n missing or a test I did not name. </context> <inputs> - Research question: [RESEARCH QUESTION] - Analysis notes: [PASTE ANALYSIS NOTES] - Missingness rule I already chose: [MISSINGNESS RULE OR UNKNOWN] - Sample n if known: [SAMPLE N OR UNKNOWN] - Software: [SOFTWARE] </inputs> <task> Write a missingness subsection: the rule I named, what I will not impute, and what I will report as unknown. If [MISSINGNESS RULE OR UNKNOWN] is UNKNOWN, list the decisions I still owe and refuse a completed plan. Leave every count as UNKNOWN unless I pasted it. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent n missing, a percent complete, or Little's test unless I named that test. - Do not invent multiple imputation if I did not name it. - If [SAMPLE N OR UNKNOWN] is UNKNOWN, refuse a missingness percent and mark UNVERIFIED. </constraints> <format> Missingness prose, then UNKNOWN cells, then UNVERIFIED rows. </format>

<context> I need the methods section to split primary analysis from exploratory work. Do not invent a pre-registered primary test I did not name. </context> <inputs> - Research question: [RESEARCH QUESTION] - Analysis notes: [PASTE ANALYSIS NOTES] - Primary analysis I named: [PRIMARY ANALYSIS] - Extra looks I already did or want: [EXPLORATORY LIST] - Registration status: [REGISTERED / NOT REGISTERED / UNKNOWN] </inputs> <task> Split the analysis: primary (only what I named), secondary if I named it, exploratory (the rest). For each extra look: keep as exploratory, drop, or UNVERIFIED if I never ran it. If registration is NOT REGISTERED or UNKNOWN, do not write that the primary test was pre-registered. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a registry ID or a locked primary outcome. - Do not move an exploratory look into primary to make the paper tidier. - If a split needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Three headed lists (primary / secondary / exploratory), then a do-not-claim line, then UNVERIFIED rows. </format>

<context> I need a software sentence that names only tools I used. Do not invent a version, a package, or a license I did not paste. </context> <inputs> - Research question: [RESEARCH QUESTION] - Software I can name: [SOFTWARE] - Versions if I have them: [VERSIONS OR UNKNOWN] - Packages or libraries if I have them: [PACKAGES OR UNKNOWN] - Analysis notes: [PASTE ANALYSIS NOTES] </inputs> <task> Write a software and environment sentence block: tools in [SOFTWARE], versions only from [VERSIONS OR UNKNOWN], packages only from [PACKAGES OR UNKNOWN]. If a version is UNKNOWN, write SOFTWARE NAMED, VERSION UNVERIFIED and list the About or sessionInfo step I must run. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent R, Python, SPSS, NVivo, or a package I did not name. - Do not invent a version number, a build, or a citation for the tool. - If you are unsure a package exists in my paste, mark UNVERIFIED, do not add it. </constraints> <format> Software sentences, then a VERSION UNVERIFIED checklist, then the lookup steps I must run. </format>

<context> I need ethics sentences that match what I am allowed to say. Do not invent an approval number, a waived review, or a board I did not name. </context> <inputs> - Research question: [RESEARCH QUESTION] - Ethics status: [ETHICS ID OR NONE / PENDING / NOT REQUIRED] - Board or office if I have it: [BOARD NAME OR NONE] - Data type: [HUMAN / ARCHIVAL / PUBLIC / OTHER] - Protocol notes: [PROTOCOL NOTES] </inputs> <task> Draft ethics facts only as I stated them: approval status, board name only if I gave it, and what the study must not claim. If status is NONE or PENDING, write that the methods section must not claim approval. If NOT REQUIRED, restate my reason from the paste, do not invent a category. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent an IRB, REB, or ethics ID. - Do not invent a waiver, an exemption letter, or a date of approval. - If a fact needs a letter I did not paste, mark it UNVERIFIED. </constraints> <format> Ethics prose, then a must-not-claim list, then UNVERIFIED rows. </format>

<context> I need consent or data-use sentences from the process I actually used. Do not invent a form, a signature, or a public-data licence I did not paste. </context> <inputs> - Research question: [RESEARCH QUESTION] - Consent or data-use notes: [PROTOCOL NOTES] - Process I used: [CONSENT / WAIVER / LICENCE / PUBLIC / OTHER] - What participants or sources were told: [WHAT THEY WERE TOLD OR UNKNOWN] - Withdrawal rule if any: [WITHDRAWAL OR NONE] </inputs> <task> Write consent or data-use notes: the process I named, what people were told only if I pasted it, withdrawal only if I named it, and what I must not claim. If [WHAT THEY WERE TOLD OR UNKNOWN] is UNKNOWN, keep that paragraph out and list the form I must open. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a consent script, a checkbox, or a GDPR clause. - Do not invent that consent was written if I only said verbal, or the reverse. - If a line needs a form I did not paste, mark it UNVERIFIED. </constraints> <format> Consent or data-use prose, then a form-to-open list, then UNVERIFIED rows. </format>

<context> I need risk and mitigation sentences that match risks I already named. Do not invent a harm, a hotline, or a closing procedure I did not paste. </context> <inputs> - Research question: [RESEARCH QUESTION] - Known risks: [KNOWN RISKS] - Mitigations I already run: [MITIGATIONS] - Population I could not reach: [MISSING POPULATION] - Protocol notes: [PROTOCOL NOTES] </inputs> <task> Draft a short risks subsection: each risk in [KNOWN RISKS], the mitigation I named, and risks I did not mitigate. Separate residual risk from ethics approval. Do not add a risk that is not in the paste just to look thorough. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a distress protocol, a counsellor, or a data-breach drill. - Do not invent that a risk was closed if I only named it. - If a mitigation needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Risk table: risk | mitigation in paste | residual | UNVERIFIED. Then 6 reusable sentences. </format>

<context> Ethics approval and study limits get mixed in my draft. Split them. Do not invent a limit-closing study or an approval I do not have. </context> <inputs> - Research question: [RESEARCH QUESTION] - Ethics status: [ETHICS ID OR NONE / PENDING / NOT REQUIRED] - Known limits: [KNOWN LIMITS] - Known risks: [KNOWN RISKS] - Protocol notes: [PROTOCOL NOTES] </inputs> <task> Write two short subsections: (1) ethics facts only as I stated them, (2) study limits I already accept. Move any sentence that is a design or sample limit out of ethics. Include the limit of using Claude on this methods draft. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent an ethics ID or a waived review. - Do not invent a limit-closing follow-up. - If a sentence needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Ethics subsection, limits subsection, then a moved-sentence list, then UNVERIFIED rows. </format>

<context> I used Claude to draft methods prose. I need a disclosure line that matches the venue policy I paste, not a hidden ghostwrite. </context> <inputs> - Research question: [RESEARCH QUESTION] - Venue AI policy I actually have: [VENUE AI POLICY] - What Claude did: [WHAT CLAUDE DID] - What I checked: [WHAT I CHECKED] - Protocol notes: [PROTOCOL NOTES] </inputs> <task> Draft 3 disclosure options: short (one sentence), methods-paragraph (3 to 5 sentences), and acknowledgements-style. Each must state that Claude drafted from my notes, that I verified procedures against the protocol, and that Claude invented no n or citations. Honor [VENUE AI POLICY]. If the policy is silent, say ASK VENUE rather than inventing a rule. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a journal rule, a COPE clause, or a required phrase I did not paste. - Do not claim I checked a step I did not list in [WHAT I CHECKED]. - If a policy line is unclear, mark UNVERIFIED and ASK VENUE. </constraints> <format> Three disclosure drafts, then ASK VENUE blanks, then UNVERIFIED rows. </format>

<context> I need a methods-section outline a paper or chapter can follow, not a finished draft. Do not stock the outline with procedures, n, or citations I have not pasted. </context> <inputs> - Research question: [RESEARCH QUESTION] - Venue headings if I have them: [VENUE HEADINGS OR NONE] - Protocol notes: [PROTOCOL NOTES] - Word count target: [WORD COUNT] - Citation style: [CITATION STYLE] </inputs> <task> Build an outline: design, setting, participants or corpus, collection, analysis, ethics, deviations, limits. Each heading needs a job, a word budget that sums to [WORD COUNT], and a blank where a count, version, or citation is missing. If I pasted [VENUE HEADINGS OR NONE], use those headings. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - If a heading has no pasted support, leave it as a heading plus NEED NOTES. - Do not invent a journal heading I did not paste. - If a claim needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Nested outline with word counts and NEED NOTES tags. UNVERIFIED list at the end. </format>

<context> I will paste the protocol notes I actually have. Draft a complete methods section from that paste only. Do not invent procedures, stats, software versions, or a literature I have not supplied. </context> <inputs> - Research question: [RESEARCH QUESTION] - Protocol notes and what I actually did: [PROTOCOL NOTES] - Word count: [WORD COUNT] - Citation style: [CITATION STYLE] - Required headings: [VENUE HEADINGS OR NONE] - Sample n if known: [SAMPLE N OR UNKNOWN] </inputs> <task> Draft the full methods section under [WORD COUNT]: design, setting, participants or corpus, collection, analysis, ethics, deviations, limits. Cite only sources present in the paste. Tag any sentence that still needs a procedure, count, version, or source as UNVERIFIED. Leave counts as UNKNOWN unless I pasted them. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not write completed results for a study that has not run. - If a heading has no pasted support, write NEED NOTES under it instead of filling it. - Prefer a thinner honest draft over a complete-looking section I cannot evidence. </constraints> <format> Full methods draft under the headings I gave (or a standard set if I said NONE), then a master UNVERIFIED and UNKNOWN list. </format>

<context> I must cut [PASTE DRAFT] to [WORD COUNT] or [PAGE LIMIT OR NONE] without losing a procedure I actually ran or inventing a tighter statistic. </context> <inputs> - Draft: [PASTE DRAFT] - Target word count: [WORD COUNT] - Page limit if any: [PAGE LIMIT OR NONE] - Must keep: [MUST KEEP] - Protocol notes: [PROTOCOL NOTES] - Citation style: [CITATION STYLE] </inputs> <task> Cut to [WORD COUNT] (and [PAGE LIMIT OR NONE] if I set one). Protect [MUST KEEP]. Prefer deleting repetition and throat-clearing. Do not delete a procedure and keep the claim as if it were still specified. Do not invent a tighter n to save space. Show a cut log. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not add new citations while cutting. - Do not merge two instruments into one name. - If a required procedure cannot survive the cut, say so and stop before you gut the protocol. - Report old and new word counts. </constraints> <format> Cut draft, word counts, cut log (what left and why), UNVERIFIED rows. </format>

<context> I need a hostile methods audit: does the draft claim more than the protocol and access can support? Do not repair an overclaim with extra data or a second coder I did not have. </context> <inputs> - Research question: [RESEARCH QUESTION] - Methods draft: [PASTE METHODS] - Protocol: [PROTOCOL] - Data I can reach: [DATA ACCESS] - Sample n if known: [SAMPLE N OR UNKNOWN] </inputs> <task> Audit [PASTE METHODS] against [PROTOCOL] for overclaims: causal language, dual coding I did not name, a representative sample I did not have, a test I cannot run, a software version I did not paste, or a completed result in past tense. For each: quote the line, name the overclaim, and rewrite a weaker line I can keep. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent n, a second coder, an instrument, or a finding. - If [SAMPLE N OR UNKNOWN] is UNKNOWN, refuse a power or representativeness claim. - If a rewrite needs a source I did not give, mark it UNVERIFIED. </constraints> <format> Table: quoted line | overclaim | weaker rewrite | UNVERIFIED? </format>

<context> I need a checklist before I submit a methods section under [VENUE RULES]. This is a QA list, not a new draft of the section. </context> <inputs> - Research question: [RESEARCH QUESTION] - Venue rules: [VENUE RULES] - Protocol notes: [PROTOCOL NOTES] - Word or page limit: [WORD COUNT] - Citation style: [CITATION STYLE] - Optional draft: [PASTE DRAFT OR NONE] </inputs> <task> Build a checklist: design match, sample honesty, collection as run, analysis without fake counts, software versions verified, ethics wording, deviations named, word count versus [WORD COUNT], citation completeness against sources I pasted, and any rule in [VENUE RULES]. Mark items I must verify by hand against the protocol. </task> <constraints> - Do not invent procedures, stats, software versions, or citations. - Do not invent a journal rule I did not paste. - Do not invent that my citations or versions are complete. - If I pasted a draft, flag only problems you can see. If I did not, keep items as blank checks. - No suggested extra papers. </constraints> <format> Checkbox list grouped by design, sample, collection, analysis, ethics, venue. Hand-verify items in italics. UNVERIFIED rows last. </format>